Close Proximity Soundings Within Supercell Environments

DECEMBER 2003THOMPSON ET AL.echoes, inflow notches, etc. (Browning 1964; Lemon1977); a WSR-88D peak cyclonic (counterclockwise) azimuthal shear of 20 m s 21 or greater at the0.5 or 1.58 elevation angles across not more than 10km [i.e., a minimum azimuthal shear of 0.002 s 21in relatively coarse 1-km resolution velocity data,similar to the Mesocyclone Detection Algorithm described in Stumpf et al. (1998)]; and persistence ofcyclonic shear for at least 30 min. All three criteriamust have been met, though only the two lowestradar elevation angles were considered due to operational data constraints at the SPC.2) Supercells were categorized as either significantlytornadic (F2 or greater tornado damage), weakly tornadic (F0–F1 tornado damage), or nontornadic. Notevery supercell was included, so that our datasetwould not be overly influenced by single days withlarge numbers of supercells. Instead, we collectedan average of roughly two cases for each day wheresupercells were identified (413 supercells from 226different days). All supercells of the same type wereseparated by at least 3 h and 185 km when multiplestorms were collected during a single day.3) Finally, a RUC-2 analysis gridpoint sounding wasgenerated for each supercell at the analysis time closest to the most intense tornadoes with the tornadicsupercells, or the time of the most intense severeweather reports with the nontornadic supercells, orat the time of the most pronounced radar signaturesif no severe weather was reported. The RUC-2 analysis soundings were interpolated (bilinear betweennearest four grid points) for each supercell to theclosest surface observing site that was generally located upwind from the supercell at the surface, perregional observations. Surface observing sites werean option for generating soundings via the NSHARPsoftware (Hart and Korotky 1991), which allowedrelatively simple identification of each case. In aneffort to eliminate so-called elevated supercells(those clearly rooted well above the surface), surface-based CAPE must have been present in eachRUC-2 proximity sounding. The RUC-2 analysisgrids were available at 40-km horizontal grid spacing, on isobaric surfaces with 25-hPa vertical resolution (e.g., 1000, 975, 950, 925 hPa, etc.) Use ofthe isobaric data resulted in a loss of vertical resolution near the ground (i.e., roughly nine levels inthe lowest kilometer of the native hybrid sigma-isentropic coordinate system, versus four levels in theisobaric grids through the same depth). Also, contamination of the RUC-2 analysis soundings by observed and parameterized convection was limited bythe relatively sparse coverage of convective precipitation in our cases. The net result was proximitysoundings that were generally within 30 min and 40km of each supercell.Following these guidelines, a nationwide sample of1245413 right-moving supercells and associated RUC-2model analysis soundings was gathered for the periodfrom April 1999 through June 2001 (Fig. 1a). Includedin this sample were 54 significantly tornadic supercells,144 weakly tornadic supercells, and 215 nontornadicsupercells. When any of these supercells occurred within3 h of a standard sounding time (0000 or 1200 UTC,or special soundings at 0600 and 1800 UTC), the nearestobserved sounding was also archived if it had 1) surfacebased parcel CAPE (Doswell and Rasmussen 1994), 2)complete data below the equilibrium level, and 3) noobvious thermodynamic or kinematic alteration by nearby thunderstorms. Finally, for each observed soundingmeeting these criteria, a RUC-2 analysis sounding validat the time and location of the observed sounding wasgenerated to determine how accurately the RUC-2 depicted the regional supercell environment (Fig. 1c).The geographic distribution of the supercell proximity soundings in Fig. 1a indicates that the vast majority of the observed supercells occurred east of theRocky Mountains, especially across the the Great Plainsand Midwest. It is important to note that not all supercells were collected for each day, and the authors wereunable to save storm information on a few occasions.For example, only two proximity soundings were gathered during the 3 May 1999 tornado outbreak in theplains that consisted of at least 10 supercells, while onlyone sounding was collected on most supercell days ina particular region. If multiple supercell types occurredin the same region (e.g., significantly tornadic and nontornadic), a RUC-2 sounding was generated for the mostintense storm of each type per the criteria discussed in3). Therefore, Fig. 1a is representative of the numberof supercell days in a particular region. Also of note inFig. 1a is a lack of events along the immediate coastsof the Gulf of Mexico and Atlantic Ocean. This wasbecause all RUC-2 proximity soundings with surfacepressures of 1000 hPa or greater were truncated erroneously to 975 hPa by a limitation in the NSHARPsounding analysis software.1 Approximately 100 lowelevation supercells, as well as more than 100 nonsupercell storms, were excluded from this study since thelowest 25–40 hPa of each RUC-2 sounding were missing. The hourly sounding distribution (not shown) reveals an expected diurnal peak near 0000 UTC, withthe vast majority of the proximity soundings confinedto the period from 1800 to 0600 UTC. More variabilityis noted in the monthly distribution of cases (Fig. 1b),where significantly tornadic storms were more frequentduring the spring, while the nontornadic and nonsupercell storms were more common in the late spring andsummer months. The relative minima in June for nontornadic and nonsupercell storms can be attributed totime away from the SPC by the lead authors, where timeconstraints necessitated collection of the less frequent1The truncation error affected only the model grid soundings, notthe observed soundings.

1246WEATHER AND FORECASTINGVOLUME 18FIG. 1. (a) Geographic plot of 413 RUC-2 supercell proximitysounding locations. Solid circles represent significantly tornadic supercells (F2–F5 tornado damage, 54 cases), open circles denote weakly tornadic supercells (F0–F1 tornado damage, 144 cases), and Xsmark nontornadic supercells (215 cases), (b) number of cases bymonth for each group, and (c) number of 0-h RUC-2 analysis soundings and observed sounding comparisons at each site (1-h forecastsounding numbers in parentheses).tornadic storms at the expense of nontornadic and nonsupercell events.A primary advantage of the RUC-2 analysis soundings is their hourly availability. The RUC-2 analysescontain asynoptic data from wind profilers, aircraft temperatures, and winds (automated weather reports fromcommercial aircraft, ACARS); WSR-88D velocity azimuth display winds; satellite winds; surface observingnetworks; etc. However, the quality of these soundingscan be questioned given the nonuniform observationsof temperature, moisture, and winds above the surfaceand between the standard twice-daily soundings at 0000and 1200 UTC. To examine the accuracy of these asynoptic RUC-2 soundings, we chose 1-h forecasts from2300 UTC, valid at 0000 UTC, to compare to the 0000UTC observed soundings. Our assumption was that the1-h forecast, based on a RUC-2 analysis 11 h after thetime of the previous synoptic soundings, should havebeen the least accurate of the day. The geographic distribution of the comparison cases is shown in Fig. 1c,which closely resembles the storm cases in Fig. 1a.Sounding errors were computed by taking either thedifference between the analysis and observed value, orbetween the forecast and observed value. Hence, pos-

DECEMBER 2003THOMPSON ET AL.1247FIG. 1. (Continued )itive (negative) errors mean that the analysis or forecastvalue was greater (less) than the observed value. Tofacilitate a consistent comparison of sounding characteristics, the observed comparison soundings wereinterpolated to the same 25-hPa pressure surfacesavailable in the RUC-2 analyses. Confidence intervalsabout the mean error were computed based on the tstatistic (Wilks 1995). The error distributions do notdeviate grossly from a normal distribution, though theerror distribution tails tend to be slightly larger thanwhat is expected from normally distributed errors.Therefore, our 95% confidence intervals may be somewhat narrow.Errors in bulk properties (such as CAPE and vectorshear magnitude) were computed in the same manneras the basic sounding variables (such as temperature).However, these errors are clearly not normally distributed. The error distribution tends to be a function of theparameter in question, and it may be partly due to thenature of some parameters. For example, CAPE cannotbe negative. Hence, the median was used to estimatethe overall error because it is resistant to outliers.3. RUC-2 sounding error characteristicsa. Temperature, mixing ratio, and wind speed errorsComparison soundings were collected for 149 RUC2 analysis (0 h) soundings (Fig. 1c), and a subset2 (125)of these analysis soundings also included the 1-h forecast soundings valid at the same time and location. The1-h forecast soundings were not collected routinely during the spring of 1999, thus the smaller sample size. Aspart of the comparison, all observed soundings wereinterpolated to the same 25-hPa isobaric surfaces available in the RUC-2 analysis grids. The profile of temperature errors for the 0-h RUC-2 analysis soundings(Fig. 2a) reveals that the zero error was generally withinthe 95% confidence interval from about 850 to 400 hPa.Temperature errors were larger near the ground, with astrong tendency for model surface temperatures to beabout 0.58C too cool. Vertical temperature errors for the2Initial data collection during 1999 included primarily the 0-hanalysis soundings, while 1-h forecast soundings were not archivedby the authors until early 2000.

1248WEATHER AND FORECASTINGVOLUME 18FIG. 2. Vertical profiles of the 95% confidence intervals for (a)temperature errors (8C), (b) mixing ratio errors (g kg 21 ), and (c) windspeed errors (m s 21 ) for 149 0-h RUC-2 analysis soundings (shaded),and a subset of 125 1-h forecast soundings valid at the analysis time(hatched).1-h RUC-2 forecast soundings (Fig. 2a) were substantially different than for the 0-h soundings, with pronounced overforecasts (roughly 0.58C) from the surfaceto 800 hPa. Schwartz and Benjamin (2002b) presentedevidence of a roughly 18C cold bias in RUC 3-h forecasts of surface (2 m) temperatures at many airport hublocations across the central and eastern United Statesduring January 2002. Model terrain elevation and landuse specifications introduced the largest errors at airportlocations in mountainous areas and near coastlines intheir sample, though our supercell cases occurred primarily across the Great Plains and Midwest where sucherrors tended to be smaller. The warm bias in our 1-hsoundings may reflect a forecast bias of the RUC-2 model itself during the warm season, which is manifestthrough a series of short-term RUC-2 forecasts that arewell removed in time from the previous synoptic soundings (e.g., the RUC-2 analyses from about 1800 to 2300UTC). Still, the temperatures errors were not particularly large and were of similar magnitude to the measurement accuracy of the radiosonde observations(0.58C; NOAA 2003).Mixing ratio errors (Fig. 2b) were largest near theground with a tendency for the RUC-2 analyses to overestimate the mixing ratios by 0.1–0.2 g kg 21 immediately above the surface. The small errors above 400 hPaare somewhat misleading since mixing ratios aloft tendto be limited by cold temperatures; thus, absolute errormagnitudes are necessarily small. Dewpoint temperatureerrors (not shown) were largest from 400 to 100 hPa.

DECEMBER 20031249THOMPSON ET AL.In summary, the 0-h RUC-2 analyses tended to be alittle too cool and dry at the surface, while the 1-hforecast soundings were too warm and moist around 900hPa. Both the 0-h analyses and 1-h forecast soundingstended to overestimate tropospheric wind speeds by 1–2 m s 21 . Contrary to our working hypothesis, the 1-hforecast error magnitudes were not substantially largerthan errors calculated from the 0-h analysis soundings,and the asynoptic soundings appear to be sufficientlyaccurate to allow their inclusion in our proximity sounding sample. It is important to note that the presentederror characteristics may represent a best-case scenario,and larger errors are possible away from the radiosondesites. Of greater concern are the impacts of these basicvariable errors on derived convective parameters.FIG. 3. Box and whiskers plot of SBCAPE and MLCAPE errors(J kg 21 ) for both the 0-h analysis and 1-h forecast soundings. Theshaded box encloses the 25th–75th error percentiles (interquartilerange), and the dark band within the interquartile range marks the95% confidence intervals for the median error values. The whiskersextend to the closest error value that is not more than 1.5 times theinterquartile range (length of the box), and the extreme errors aremarked by the solid dots above and below the whiskers.The 1-h forecast mixing ratios were generally 0.2–0.3g kg 21 too large near 900 hPa (Fig. 2b), though surfacevalues were too low by roughly the same amount. Recent work by Turner et al. (2003) documented a 5% drybias in the radiosondes used in National Weather Service(NWS) field offices, which is of similar magnitude tothe mean RUC-2 analysis errors in this sample, as wellas the relative humidity accuracy of the observed soundings (5%; NOAA 2003).The RUC-2 analysis wind speed tended to be about1–2 m s 21 too strong from the surface to 400 hPa (Fig.2c). Despite the speed overestimates in the RUC-2, theprofiles of u and y wind component errors (not shown)for the 0-h analysis soundings were consistent with the1-h forecasts, with the zero error generally within the95% confidence interval. However, there was someskew in the profiles such that the zonal (westerly) andmeridional (southerly) wind components were overforecast by 0.5–1 m s 21 in the layer from the surfaceto about 600 hPa. Our surface wind speed errors aresimilar to the findings of Schwartz and Benjamin(2002b), and are of similar magnitude to the measurement accuracy of the radiosondes (1.5 m s 21 ; NOAA2003).b. Bulk parameter errorsSeveral bulk supercell sounding parameters related toCAPE and vertical shear were also examined for boththe analysis and 1-h forecast soundings. In general, thesurface-based CAPE (SBCAPE) values were underestimated by 300–500 J kg 21 in the 0-h soundings (seeFig. 3 and Table 1), due to the negative biases in surfacetemperatures and mixing ratios (see Fig. 2a). The 1-hforecast sounding errors for SBCAPE covered a rangeof values similar to the 0-h sounding errors, though theerrors were more closely centered near zero as a resultof minor warm and dry biases counteracting one anotherin the CAPE calculation. The 0-h 100-mb mean parcelCAPE (MLCAPE) errors were less biased than theSBCAPE errors, with a tight clustering of values insmall range (0 to 2250 J kg 21 ).A closer examination of the extreme outlier SBCAPEerror (23930 J kg 21 ) revealed that a near-surface‘‘spike’’ in the dewpoint temperature was responsiblefor the difference between the observed and RUC-2analysis soundings (Fig. 4). The MLCAPE error(;2240 J kg 21 ) for these same soundings shown inFig. 4 was much smaller than the SBCAPE error. Theobserved moisture profile in Fig. 4 can be questionedgiven the surface mixed-layer depth in excess of 100hPa, though no additional observations were availableto dispute the rapid moisture decrease just above thesurface. Schwartz and Benjamin (2002a) discussed various choices in CAPE calculations and documented 3h RUC errors similar to our MLCAPE errors, and Craven et al. (2002b) found that a 100-hPa mean parcelTABLE 1. Mean parameter values (mean), mean absolute errors (MAEs), and mean arithmetic errors (bias) for the 0- and 1-h RUC-2comparison soundings.Parameter0-h mean0-h MAEs0-h bias1-h mean1-h MAEs1-h biasSBCAPE (J kg21 )MLCAPE (J kg21 )MLLCL (m AGL)0–1-km shear (m s21 )0–6-km shear (m s21 .2

1250WEATHER AND FORECASTINGVOLUME 18FIG. 5. Same as in Fig. 3 except for 0–1- and 0–6-km vector shearmagnitude error (m s 21 ).FIG. 4. Skew T2logp plot of the RUC-2 analysis sounding (black)for the extreme SBCAPE error from Fig. 3, with an overlay of thecollocated observed sounding (gray).was superior to a surface parcel in determining parcelascent. As discussed in the previous subsection, RUCforecast errors in surface variables can be magnified bydifferences between observed and model terrain height,and in regions of sharp transition in land use or surfacetype (i.e., coastal regions). Considering the findings ofthe aforementioned studies, and the error characteristicsof our comparison soundings, the use of a mean (100hPa) layer parcel appears to be a better choice than asurface parcel in evaluating environmental characteristics of supercell environments with RUC model proximity soundings.The RUC-2 tended to overforecast the 1-h MLCAPEas a result of temperatures and mixing ratios being toolarge within the lowest 100 hPa (cf. Figs. 2a and 2bwith Fig. 3). An examination of individual comparisoncases suggests that the warm and moist bias in the 1-hforecast soundings may be attributable to the RUC-2generating a ‘‘well mixed’’ boundary layer that is toodeep, though the specific sources of these errors in themodel are unknown. Despite the apparent differencesin MLCAPE error characteristics between the 0- and 1h forecast soundings, experience of the authors suggeststhat these error magnitudes are too small to have a serious impact on operational evaluation of storm environments. We believe the MLCAPE values to be lessbiased and more representative of the potential for deepconvection than SBCAPE; hence, we have chosenMLCAPE for our proximity sounding evaluation. Similarly, mean layer lifting condesation level (MLLCL)errors (Table 1) were considered to be reasonably smalland not biased in the mean, while error distributions(not shown) were centered near zero for both the 0- and1-h soundings.Errors in vertical shear parameters, such as stormrelative helicity3 (SRH) and measures of deeper layershear (e.g., 0–6-km vector shear magnitude), were moreuniformly distributed than the CAPE errors. The 95%confidence intervals for median values of 0–1- and 0–6-km vector shear magnitude errors (Fig. 5) are nearzero error in each layer for both the 0-h analyses and1-h forecasts, with typical error magnitudes around 2m s 21 and small mean errors (see Table 1). Similar errordistributions were also documented for the 0–1- and 0–3-km SRH (not shown). In spite of the tendency for theRUC-2 analyses and 1-h forecasts to overestimate windspeeds throughout the troposphere in our comparisonsoundings, the derived shear parameters were not biasedbecause the speed and direction errors were relativelyconsistent throughout the lowest 6 km (i.e., the surfaceto about 500 hPa). These results suggest that the ourRUC-2 proximity soundings were reasonably representative of the regional supercell environments, and canbe used to compare parameter distributions across thestorm groupings described in section 2.4. Supercell and tornado forecast parametersThe 413 supercell proximity soundings were categorized as associated with either 1) nontornadic (215cases), 2) weakly tornadic (F0–F1 tornado damage, 144cases), or 3) significantly tornadic (F2–F5 tornado damage, 54 cases) supercells. In addition, a sample of 753SRH values were calculated using storm motions estimated bythe algorithm developed by Bunkers et al. (2000), since these comparison soundings were not considered to be direct proximity soundings for any particular supercell. The Bunkers algorithm is the mostreliable means to estimate supercell motion that is now available inforecast operations, as independently verified by Edwards et al.(2002).

DECEMBER 2003THOMPSON ET AL.1251FIG. 6. Box and whiskers plot of MLCAPE values (J kg 21 ) with the significantly tornadicsupercells (sigtor, 54 soundings), weakly tornadic supercells (weaktor, 144 soundings), nontornadic supercells (nontor, 215 soundings), ‘‘marginal’’ supercells (mr

DECEMBER 2003 THOMPSON ET AL. 1243 q 2003 American Meteorological Society Close Proximity Soundings within Supercell Environments Obtained from the Rapid Update Cycle RICHARD L. THOMPSON,ROGER EDWARDS, AND JOHN A. HART Storm Prediction Center, Norman, Oklahoma